Evaluation of indigenous and exotic wheat genotypes for tolerance to severe terminal heat stress in the agro-climatic conditions of Bihar

1 Introduction

Wheat (Triticum aestivum L.), a major cereal crop of the Poaceae family, is a primary source of carbohydrates and calories globally, meeting the dietary needs of over one-third of the population (Liu et al., 2020; Khalid et al., 2023). Originating in the Fertile Crescent (Feldman and Levy, 2015; Heun et al., 1997), wheat is especially vulnerable to rising global temperatures with terminal heat stress emerging as a major constraint during the grain-filling stage. Terminal heat stress severely affects biochemical and physiological processes in wheat reducing photosynthesis, pollen viability, grain development and ultimately grain yield (Prasad et al., 2017; Riaz et al., 2021). High temperatures experienced during the reproductive and grain-filling stages are known to accelerate senescence and reduce the duration of grain filling (Kumar H. et al., 2023; Fischer and Wood, 1979). Under late sowing crops often face 25–32 °C during anthesis and grain filling leading to early maturation and substantial yield loss (Lamba et al., 2023). Studies worldwide reveal a 3–17% wheat yield reduction each 1 °C increases in average temperature across regions like South Asia, China, Mexico and the U.S (Lobell et al., 2005; Tiwari et al., 2013). In extreme scenarios, yields may decline up to 50% (Asseng et al., 2011). Approximately 0.8–1.2 million hectares in Bihar are exposed to terminal heat stress, where temperatures during grain filling frequently exceed 31 °C, resulting in significant yield penalties (Sattar et al., 2020). These projections highlight the urgent need to develop heat-tolerant wheat genotypes. Despite growing awareness comparative studies between indigenous and exotic genotypes remain limited. Most research evaluates local or introduced germplasm separately hindering insights into their relative adaptive responses to terminal heat stress. Additionally, proposed heat tolerance mechanisms like sustained photosynthesis and efficient stem reserve remobilization lack consistent validation across diverse environments reducing their breeding utility. Furthermore, prolonged heat stress effects on reproductive traits like grain-filling duration and physiological maturity are underexplored, especially in exotic genotypes. While indigenous genotypes may show delayed senescence or stay-green traits their comparative advantages remain poorly documented (Reynolds et al., 2021). Optimizing genotype-specific agronomic practices such as sowing dates and nutrient management is another underdeveloped area.

Recent studies highlight significant variability in stress responses across indigenous and exotic genotypes evident in traits like chlorophyll content, proline accumulation, and antioxidant enzyme activity (Dhyani et al., 2013). Heat tolerant lines tend to maintain higher chlorophyll levels, delayed senescence and robust antioxidative defenses under stress, contributing to greater yield stability. Field-based evaluations using integrated phenotypic and physiological screening have proven effective for identifying superior genotypes under heat stress (Fu et al., 2023; Gautam et al., 2023). Incorporating both indigenous and exotic germplasm expands the genetic base and enhances breeding strategies for climate-resilient wheat (Yuling Zheng et al., 2025). While advanced methods like mutagenesis, synthetic wheat development and genetic engineering have contributed to heat tolerance research (Mullarkey and Jones, 2000; Rattey et al., 2011; Fu et al., 2008), conventional breeding remains the cornerstone for developing tolerant cultivars through strategic screening and selection. The field experiment reported in this study forms part of a broader research programme on terminal heat stress tolerance in bread wheat. An earlier publication from this experiment presented the first-year (2022–23) phenotypic assessment, focusing on trait relationships through correlation and path coefficient analyses (Kumar et al., 2026). The present study extends that work by integrating phenotypic data from two consecutive cropping seasons (2022–23 and 2023–24) and employing pooled analyses, BLUP-based evaluation, genetic parameter estimation, stress tolerance indices, and principal component analysis to identify superior heat-tolerant wheat genotypes. Building on these insights, the present study aimed to (i) evaluate the performance of indigenous and exotic wheat genotypes under contrasting terminal heat stress environments using phenotypic data collected over two consecutive cropping seasons (ii) estimate genetic parameters and calculate stress tolerance indices to characterize genotype responses to terminal heat stress; and (iii) identify superior wheat genotypes with enhanced terminal heat stress tolerance for their potential use in wheat breeding programs.

2 Materials and methods

2.1 Experimental site

The research was carried out during the rabi seasons of 2022–23 and 2023–24 at the experimental field of Dr. Rajendra Prasad Central Agricultural University, located in Pusa, Samastipur, Bihar. The site features deep, calcareous alluvial soils with textures ranging from light to heavy. The experimental material consisted of 500 wheat genotypes comprising indigenous and exotic germplasm. The exotic entries were primarily from CIMMYT international nurseries, including the Semi-Arid Wheat Screening Nursery (SAWSN), High temperature Wheat Yield Trial (HTWYT), Elite Spring Wheat Yield Trial (ESWYT), Harvest Plus Yield Trial (HPYT) and Helminthosporium Leaf Blight Screening Nursery (HLBSN). These materials developed and selected for adaptation to heat- and moisture-stressed environments, disease resistance, and yield potential. Indigenous entries consisted of released varieties and advanced breeding lines maintained at the Department of Genetics and Plant Breeding, RPCAU, Pusa. The germplasm panel was assembled to ensure broad genetic diversity for evaluating terminal heat stress tolerance under field conditions. The list of exotics and indigenous germplasm presented in Supplementary Table 1.

2.2 Experimental design

The genotypes were evaluated under three sowing conditions: timely (TS), late sown (LS) and very late sown (VLS), with sowing dates of November 15, December 9 and January 3, respectively. Seeds were sown directly into the field under adequate soil moisture conditions. The experiment followed an alpha lattice design with three replications. Each replication consisted of 20 incomplete blocks, with each block comprising 25 entries. Each plot measured 3 m² and consisted of five rows each 3 meters long with 20 cm spacing between rows. Replication and incomplete block effects were included in the ANOVA model to account for field heterogeneity and local spatial variation, thereby improving the precision of genotype evaluation. To support crop growth three irrigations were applied at crown root initiation, late jointing and milking stages. As the soil has good water holding capacity, no additional irrigation was required for any TS, LS and VLS conditions. Fertilizers were applied as per recommended doses 120 kg/ha nitrogen (split between pre-sowing and pre-flowering), 60 kg/ha phosphorus and 40 kg/ha potassium. Standard agronomic practices were followed and no major pest infestations were observed during the study. Genotypes were harvested at physiological maturity.

2.3 Sampling and measurements

In order to take non-destructive measurements, five plants per genotype were chosen and tagged from the middle rows of each replication. Chlorophyll content of the flag leaf was measured using a self-calibrating SPAD-502 chlorophyll meter (Konica Minolta Sensing Inc., Tokyo, Japan) on randomly selected plants. Readings were taken on the flag leaf at 50% heading and during the critical grain-filling stage between 09:00 and 11:30 hours. Canopy temperature was measured during days to 50% heading and the grain-filling stage using a handheld infrared radiometer (MI-210, Apogee Instruments, USA). Measurements were taken at a 30° angle from the horizontal and perpendicular to the crop rows between 11:00 AM and 1:00 PM to minimize diurnal fluctuations. Days to maturity were recorded as the number of days from sowing until more than 80% of the spikes turned golden yellow. Days to 50% heading were noted as the number of days from sowing until half of the plants in a genotype reached anthesis. Yield-related traits includes spikelet fertility percentage was calculated by dividing the number of fertile spikelets by the total number of spikelets per spike, grains per spikelet were determined by dividing the total grain count per spike by the number of fertile spikelets, grain filling duration (GFD) was calculated as the number of days between days to 50% heading and physiological maturity, thousand-grain weight (g) was recorded by averaging the weight of 1000 randomly selected grains per replicate. The number of grains per spike was obtained by counting the total grains on the main spike at harvest. Grain yield per plant (g) was assessed by weighing the total grains from each individual plant (Table 1).

Table 1. The details of observation recorded during experiment.

2.4 Statistical analysis

Phenotypic data were analyzed using R software (version 4.4.0). In this study, the term environment refers to the three sowing conditions, namely timely sown (TS), late sown (LS), and very late sown (VLS), imposed to generate contrasting levels of terminal heat stress.

A pooled analysis of variance (ANOVA) was performed across two years (2022 and 2023) and three sowing environments using an alpha-lattice design. The following linear mixed model was fitted for the combined analysis across years and sowing environments:

Yijklmn=μ+Yi+Ej+(YE)ij+Rk(ij)+Bl(kij)+Gm+(GY)mi+(GE)mj+(GYE)mij+ϵijklmn

where Y_ijklmn is the observed phenotypic value; μ is the overall mean; Y_i​ is the fixed effect of the i^th year; E_j​ is the fixed effect of the j^th sowing environment; (YE)_ij​ is the fixed year × environment interaction; R_k(ij)​ is the random effect of the k^th replication nested within year × environment; B_l(kij)​ is the random effect of the incomplete block nested within replication, year, and environment; G_m​ is the random effect of the m^th genotype; (GY)_mi​, (GE)_mj​, and (GYE)_mij​ are the random genotype × year, genotype × environment, and genotype × year × environment interaction effects, respectively; and ϵ_ijklmn​ is the residual error. Significance of fixed effects was assessed using F-tests from the pooled ANOVA, and statistical significance was declared at P < 0.01.

For estimation of best linear unbiased predictions (BLUPs), separate mixed models were fitted independently for each sowing environment (TS, LS, and VLS) by combining data from both experimental years. Thus, a separate model was fitted for each environment:

Yijkl= μ+Yi+Gj+(GY)ij+Rk(i)+Bl(ki)+ϵijkl

where Y_i​ (year) was treated as a fixed effect, whereas genotype (G_j​), genotype × year interaction (GY)_ij​, replication within year R_k(i)​, incomplete blocks nested within replication and year B_l(ki)​ were treated as random effects and ϵ_ijklis the residual error. Variance components were estimated using restricted maximum likelihood (REML) implemented in the “lme4’’ package in R (version 4.4.0). Genotypic BLUPs were obtained as the sum of the overall intercept and the predicted random genotypic effect.

Broad-sense heritability was estimated from the variance components obtained from the mixed model according to:

H2=σ2gσ2g+σ2gyy+σ2gee+σ2gyeey+σ2erey

where σ²g, σ²gy​, σ²ge​, σ²gye, and σ²e​ represent the genotypic, genotype × year, genotype × environment, genotype × environment × year, and residual variance components, respectively; and y, e, and r denote the number of years, environments, and replications.

Variance components were estimated from the mixed-effects model using restricted maximum likelihood (REML), and the resulting best linear unbiased predictions (BLUPs) were subsequently used for correlation analysis and principal component analysis (PCA) (Hotelling, 1933). Correlation analysis among traits was performed using the “metan’’ package in R. Principal component analysis was carried out using the “FactoMineR’’, “factoextra’’ and “ggplot2’’ packages in R (version 4.3.2). Eigenvalues, the percentage of explained variance, and trait loading coefficients were extracted to identify the major sources of variation among genotypes. Principal component biplots and trait contribution plots were generated using the “factoextra’’ and “ggplot2’’ packages. These analyses provided a comprehensive understanding of the relationships among physiological and yield-related traits and facilitated the identification of superior genotypes under terminal heat stress.

Stress tolerance and susceptibility indices, namely stress susceptibility index (SSI), tolerance index (TOL), mean productivity (MP), stress tolerance index (STI), geometric mean productivity (GMP), yield index (YI), and yield stability index (YSI), were calculated for each genotype using grain yield under timely-sown (Yp, non-stress) and very late-sown (Ys, heat-stress) conditions. The equations and references used for the calculation of these indices are presented in Table 2. These indices were used to assess genotype performance and identify heat-tolerant genotypes (Bidinger et al., 1987).

Table 2. Different indices used to assess the performance of genotypes.

3 Results

3.1 Environmental conditions

Significant differences in meteorological conditions were observed among sowing environments and crop seasons (Figure 1);. Based on the mean days to 50% heading, anthesis occurred during 7–13 February, 14–20 February, and 28 February–6 March under TS, LS, and VLS conditions, respectively. The corresponding mean maximum/minimum temperatures during anthesis increased progressively from 24.65/9.17 °C under TS to 26.19/11.43 °C under LS and 29.02/14.17 °C under VLS, indicating greater exposure to elevated temperatures under delayed sowing. (Supplementary Table 6b) The heat stress environment was further characterized using temperature records during grain filling (Supplementary Table 6a; Figure 1). In 2022–23, mean maximum temperature during grain filling increased from 31.03 °C in TS to 32.10 °C in LS and 35.71 °C in VLS. Correspondingly, the number of days with maximum temperature exceeding 30 °C increased from 16 days in TS to 21 days in LS and 25 days in VLS. In 2023–24, mean maximum temperature during grain filling was 31.67 °C in both TS and LS, whereas VLS experienced a substantially higher temperature of 36.44 °C. The corresponding number of days with temperatures exceeding 30 °C was 6, 6, and 25 days for TS, LS, and VLS, respectively. These results indicate that VLS conditions imposed both greater heat intensity and longer heat stress duration during grain filling, confirming the successful establishment of terminal heat stress environments. Under the very late-sown environment, the crop was exposed to elevated temperatures during grain filling, with maximum temperatures exceeding 30 °C, a threshold known to adversely affect grain filling, accelerate senescence, and reduce grain yield in wheat. The severity of terminal heat stress was further reflected by substantial reductions in grain yield under the very late-sown environment, with average yield losses of 44.90% and 46.01% among the indigenous and exotic genotypes, respectively, and up to 54.03% in the susceptible check RG-1 (Supplementary Table 4). These temperature conditions and associated yield losses confirmed that the very late-sown environment imposed severe terminal heat stress.

3.1.1 Phenotypic variability and frequency distribution across sowing regimes

Before genotype selection, the extent and distribution of phenotypic variability were examined across timely sowing (TS), late sowing (LS), and very late sowing (VLS) conditions. Frequency distribution curves for grain yield and associated traits revealed a progressive shift in trait expression with increasing terminal heat stress (Figure 2). Under TS, trait distributions were centered around higher mean values with relatively narrow spreads, reflecting favorable growing conditions. In contrast, LS and particularly VLS conditions exhibited a pronounced shift toward lower mean values, indicating the adverse impact of terminal heat stress on phenotypic performance. Notably, frequency curves under VLS showed wider dispersion compared to TS, suggesting differential genotypic responses under severe heat stress. This wide range of responses confirms both the severity of terminal heat stress and the availability of diverse adaptive mechanisms among genotypes, providing a strong basis for large-scale screening.

3.2 Pooled analysis of variance, variance components, and heritability

To determine whether the observed variability across sowing regimes was genetically controlled, pooled analysis of variance (ANOVA) was performed for grain yield and associated physiological traits across TS, LS, and VLS environments (Table 3). The pooled ANOVA revealed highly significant (P ≤ 0.01) genotype effects for all major traits, indicating substantial genetic variability within the association panel (Gezahegn et al., 2015). This confirms that the wide phenotypic spread observed in frequency distributions was not merely due to environmental noise.

Table 3. Pooled ANOVA showing effects of genotype, environment, and G×E under severe terminal heat stress conditions.

SPADH, SPAD at heading; SPADM, SPAD at maturity; CT, canopy temperature; DFH, days to 50% heading; SF, spikelet fertility; GFD, grain filling duration; DPM, days to physiological maturity; TGW, thousand grain weight; NGS, number of grains per spike; GYP, grain yield per plant (*p < 0.05; p < 0.01 and *p < 0.001).

Variance component analysis revealed considerable differences among traits in the relative contributions of genotypic and interaction variance components (Table 4). Genotypic variance (σ^2g​) was highest for SPAD chlorophyll content at heading (SPADH; 1.5051), followed by days to physiological maturity (DPM; 1.1503), spike fertility (SF; 0.6694), SPAD chlorophyll content at maturity (SPADM; 0.6035), and grain yield per plant (GYP; 0.5283), indicating the presence of genetic variability for these traits. In contrast, genotype × year (σ^2GY​) interaction variance exceeded the genotypic variance for most traits and was particularly high for number of grains per spike (NGS), spike fertility (SF), SPAD chlorophyll content at heading (SPADH), and days to 50% heading (DFH), demonstrating a strong influence of seasonal variation on genotype performance. Genotype × environment (σ^2GE​) and genotype × environment × year (σ^2GEY​) interaction variances also contributed substantially to the total phenotypic variation, particularly for NGS, DPM, TGW, SF, and SPADM, indicating differential genotype responses under contrasting sowing environments across years. Variance component estimates calculated separately for each sowing environment (TS, LS, and VLS) are presented in Supplementary Table 6c.

Table 4. Variance components and broad-sense heritability across the environment.

Broad-sense heritability estimates ranged from 0.0000 to 0.2524 (Table 4). The highest heritability was observed for days to physiological maturity (DPM; 0.2524), followed by canopy temperature (CT; 0.2056), grain yield per plant (GYP; 0.1517), SPAD chlorophyll content at heading (SPADH; 0.1457), and SPAD chlorophyll content at maturity (SPADM; 0.0943). Lower heritability estimates were obtained for spike fertility (SF; 0.0437), days to 50% heading (DFH; 0.0091), grain filling duration (GFD; 0.0033), thousand-grain weight (TGW; 0.0018), and number of grains per spike (NGS; 0.0000). The relatively low heritability estimates indicate that a substantial proportion of the phenotypic variation was attributable to genotype × year, genotype × environment, and genotype × environment × year interactions, highlighting the importance of multi-year and multi-environment evaluation for the reliable identification of terminal heat-tolerant wheat genotypes.

In view of the importance of severe heat stress in differentiating genotype responses, the distribution and variability of individual traits were further assessed under the VLS condition (Table 5). Considerable variability was observed for most traits. SPAD chlorophyll content at heading and maturity ranged from 27.84–39.81 and 11.18–20.08, respectively, while canopy temperature varied from 25.57 to 29.91 °C. Grain yield per plant ranged from 5.67 to 14.69 g, number of grains per spike from 12.56 to 34.78, and thousand-grain weight from 16.17 to 30.39 g. The coefficient of variation was highest for grain yield per plant (17.82%), followed by number of grains per spike (17.25%), SPAD chlorophyll content at maturity (11.05%), and thousand-grain weight (10.92%), indicating substantial variability for these traits under severe heat stress. In contrast, days to physiological maturity (1.05%) and canopy temperature (2.22%) exhibited relatively low variation. The wide variation observed for grain yield and its associated traits suggests their potential utility in identifying heat-tolerant genotypes under terminal heat stress conditions.

Table 5. Traits means and variations for wheat genotypes under very late sown(VLS) condition.

3.3 Correlation analysis under severe terminal heat stress

VLS represents the most severe and discriminative environment, where trait–yield relationships most accurately reflect heat tolerance mechanisms rather than yield potential alone.

To identify traits contributing to yield expression under terminal heat stress, pearson’s correlation analysis (Figure 3) revealed that grain yield per plant (GYP) was significantly and positively correlated with spike fertility (SF; r = 0.62*), number of grains per spike (NGS; r = 0.50), thousand grain weight (TGW; r = 0.51), and SPAD chlorophyll content at heading (SPADH; r = 0.39), indicating the importance of sink size and photosynthetic capacity in yield determination (Figure 3). GYP showed significant negative correlations with days to flowering (DFH; r = −0.45) and days to physiological maturity (DPM; r = −0.26). Phenological traits were strongly interrelated, with DFH showing a high positive correlation with DPM (r = 0.68). Canopy temperature (CT) exhibited weak but significant negative correlations with grain yield per plant (r = −0.13), grain filling duration (r = −0.13*), number of grains per spike (r = −0.10), and thousand grain weight (r = −0.11*), while its associations with several other traits were non-significant.

3.4 Evaluation under very late sown condition (VLS)

3.4.1 Principal component analysis

3.4.1.1 Contribution of traits to phenotypic variation

Principal component analysis (PCA) based on standardized trait values under very late sown conditions revealed clear multivariate differentiation among genotypes the principal component analysis condensed the ten traits into four major principal components which accounted for 69.84% of the total variation (Iezzoni and Pritts, 1991). The first four principal component axis recorded eigen values greater than one whereas, the fifth and further principal components recorded value less than one (Figure 4, Table 6). Thus, those PC (<1) could be discarded to further shorten the set of data at disposal (Nachimuthu et al., 2014). PC 1 accounted for 35.85% of total variability and it was positively contributed by the traits. PC 1 had an eigenvalue of 3.5848 and explained 35.85% of the total variability (Jain and Patel, 2016; Kumar et al., 2016; Beheshtizadeh et al., 2013; Zewdu et al., 2024). It showed high positive loadings for Grain Yield/Plant (0.4272), Spikelet fertility (0.3875), Number of Grains/Spike (0.3172), SPAD (Heading) (0.3060), Grain filling duration (0.2962), SPAD (Maturity) (0.2652), Thousand Grain Weight (0.2725), with negative loadings for Days to 50% Heading (-0.4039), Days to Physiological Maturity (-0.2608), and Canopy Temperature (-0.0978). PC 2 had an eigenvalue of 1.2844 and accounted for 12.84% of the variability. This component was mainly associated with positive loadings for Days to 50% Heading (0.5429), Days to Physiological Maturity (0.4692), SPAD (Heading) (0.3638), SPAD (Maturity) (0.3485), Thousand Grain Weight (0.2547), and Number of Grains/Spike (0.2004), while Grain filling duration (-0.3066) and Canopy Temperature (-0.0027) loaded negatively or nearly zero. PC 3, with an eigenvalue of 1.0809, contributed 10.81% of the total variation. It was characterized by a strong positive loading of Canopy Temperature (0.6330) and marked negative loadings of Days to Physiological Maturity (-0.5253) and Grain filling duration (-0.5125), with only small contributions from yield traits. PC 4 had an eigenvalue of 1.0340 and explained 10.34% of the variability. It showed a high positive loading for Thousand Grain Weight (0.4837), along with moderate positives for Grain Yield/Plant (0.2035) and Number of Grains/Spike (0.1603), whereas SPAD (Heading) (-0.3571), SPAD (Maturity) (-0.4105), Grain filling duration (-0.4327), Canopy Temperature (-0.3518), and Days to Physiological Maturity (-0.2661) loaded negatively.

Table 6. Eigen values, percentage of variance explained, cumulative variance, and trait loadings for the first four principal components of wheat genotypes under very late sown (VLS) conditions (Pearson, 1901).

3.4.2 PCA biplot and genotype differentiation

The PCA biplot based on the first two principal components explained 48.69% of the total variation and revealed clear multivariate differentiation among genotypes (Figure 5). Genotypes positioned in the positive quadrant of PC1 were characterized by high grain yield, superior yield components, and favorable physiological traits, representing heat-tolerant and high-performing genotypes. In contrast, genotypes associated with negative PC1 values were linked with longer phenological duration and reduced yield expression, indicating heat sensitivity. Cluster analysis based on PCA scores further resolved traits into four distinct groups, separating canopy temperature, phenological traits, grain filling duration, and yield-associated traits into independent clusters. This multivariate structuring confirms that terminal heat tolerance is governed by coordinated expression of multiple traits rather than single traits in isolation.

3.4.3 Year-to-year reproducibility of grain yield under terminal heat stress

Grain yield under very late sown conditions exhibited a significant positive association between the two years of evaluation (Figure 6). Yield performance during 2022–23 was moderately and positively correlated with yield in 2023–24 (r = 0.52, P < 0.001), indicating reasonable degree of consistency in genotype performance across years under terminal heat stress. Although the correlation was moderate, the significant relationship suggests that a proportion of the genotypic differences observed under stress were reproducible across seasons despite variation in environmental conditions between years. The linear trend observed in the scatter plot further supports the stability of genotype responses under very late-sown conditions and justifies the use of multi-year performance as a criterion for identifying promising heat-tolerant genotypes.

3.4.4 Selection of superior genotypes for terminal heat tolerance

Principal component analysis, together with the ranking of all 500 genotypes based on individual heat stress tolerance indices (Supplementary Table 7a), identified a distinct group of genotypes occupying the positive quadrant of the biplot, characterized by favorable associations with grain yield under terminal heat stress. Following validation of year-to-year yield reproducibility, the top 10% of genotypes (50 entries) were shortlisted for further analysis (Supplementary Table 7b, c). Genotypes exhibiting consistent yield performance across both years under VLS conditions were initially identified. Within this stable set, priority was given to genotypes with lower stress susceptibility index (SSI), as lower SSI values indicate reduced sensitivity to heat stress. Subsequently, genotypes were evaluated across other stress tolerance indices, including mean productivity (MP), stress tolerance index (STI), geometric mean productivity (GMP), yield stability index (YSI), yield index (YI), and tolerance (TOL). The final selection of 22 genotypes (Table 7) was based on their ability to simultaneously maintain high grain yield under stress and exhibit favorable values across multiple indices. Thus, selection was not based solely on SSI or any single index, but rather on the combined evidence from stress tolerance indices, PCA scores, and multi-trait performance (Bhandari et al., 2024). These selected genotypes formed the basis for subsequent clustering and profiling analyses. A total of 500 wheat genotypes, comprising 119 indigenous and 381 exotic entries, were evaluated under timely-sown (TS), late (LS) and very late-sown (VLS) conditions. Based on grain yield under heat stress and stress tolerance indices (SSI, STI, GMP, MP, YI, and YSI), 25 superior entries were identified. Among these, 16 were indigenous and 9 were exotic; however, three indigenous entries (DBW 14, HD 2967, and RG-1) were included as check cultivars. After excluding the checks, 22 promising heat-tolerant genotypes were identified, consisting of 13 indigenous and 9 exotic genotypes (Supplementary Table 2).

Table 7. Stress tolerance indices and their trait values of selected genotypes along with checks.

Although indigenous germplasm represented only 23.8% of the evaluated panel, it contributed 59.1% (13 of 22) of the selected heat-tolerant genotypes. The frequency of selected genotypes differed significantly between indigenous and exotic germplasm groups (χ² = 13.83, df = 1, p < 0.001), indicating a greater enrichment of heat-tolerant genotypes within the indigenous germplasm. The performance of the selected indigenous and exotic genotypes identified from the diversity panel and is intended to provide a descriptive comparison of elite entries (Supplementary Table 3). To statistically evaluate differences between indigenous and exotic germplasm groups, Welch’s t-tests were performed using the complete diversity panel comprising all evaluated genotypes. The results indicated significant differences for SPADH and grain filling duration (GFD) after FDR correction, whereas the remaining traits did not differ significantly between the two germplasm groups (Supplementary Table 5).

The selected entries exhibited favorable combinations of high grain yield under stress, low stress susceptibility index (SSI), high stress tolerance index (STI), geometric mean productivity (GMP), mean productivity (MP), yield index (YI), and yield stability index (YSI), and were therefore considered promising sources of terminal heat tolerance (Supplementary Tables 3, 4).

3.4.5 Genotype clustering and top genotype profiling

PCA-based clustering grouped the 22 selected genotypes into two distinct clusters: Cluster II comprising heat-tolerant genotypes and Cluster I representing moderately heat susceptible genotypes (Figure 7). Clear separation of cluster centroids validated the robustness of the grouping approach. Following the classification criteria proposed by Baranwal et al. (2024), genotypes with SSI values ranging from 0.00 to 0.75 (SSI ≤ 0.75) were categorized as heat tolerant, whereas genotypes with SSI values ranging from 0.76 to 1.25 (SSI > 0.75 and ≤ 1.25) were considered moderately susceptible to terminal heat stress. Accordingly, the top nine genotypes (including one check) belonging to the tolerant group (SSI ≤ 0.75) were further profiled using radar plots, illustrating their superior and balanced performance across multiple yield and physiological traits simultaneously (Figure 8). These genotypes represent promising candidates for breeding programs aimed at improving terminal heat tolerance in wheat. To further assess their stability and adaptability, the performance of the heat-tolerant genotypes was compared across the three environments (Figure 9). Grain yield and SSI values of all 22 genotypes were subsequently examined to identify contrasting responses to terminal heat stress and validate genotype classification into tolerant and susceptible groups (Figure 10). Furthermore, the relationships between grain yield per plant and different stress tolerance indices were explored to evaluate the effectiveness of these indices in identifying heat-tolerant genotypes (Figure 11).

Comparative analysis of the selected indigenous and exotic germplasm revealed distinct differences in their response to terminal heat stress. Indigenous genotypes exhibited slightly higher grain yield under heat stress (Ys = 13.63 g plant^-¹) than exotic genotypes (13.57 g plant^-¹), despite having a lower yield potential under timely-sown conditions (Yp = 24.81 vs. 25.19 g plant^-¹). Indigenous entries also recorded lower mean yield reduction under VLS conditions (44.90%) compared with exotic entries (46.01%), along with lower stress susceptibility (SSI = 0.76 vs. 0.78) and marginally higher yield stability (YSI = 0.55 vs. 0.54). In addition, indigenous genotypes showed higher spike fertility and thousand-grain weight, indicating a superior capacity to sustain reproductive growth and grain filling under heat stress.

The mean performance of both indigenous and exotic groups was superior to that of the three check cultivars (DBW 14, HD 2967, and RG-1). Both groups exhibited higher grain yield under stress (Ys = 13.63 and 13.57 g plant^-¹, respectively) than the checks. Likewise, mean STI (0.56) and GMP (18.38–18.48) values of the selected germplasm exceeded those of the checks, indicating greater productivity and stability across stress and non-stress environments. Among the checks, HD 2967 displayed comparatively better stability (SSI = 0.68; YSI = 0.60) but lower grain yield under stress, whereas RG-1 was the most susceptible, exhibiting the highest yield reduction (54.03%) and lowest YSI (0.46). These results demonstrate that the selected genotypes surpassed the standard checks for several key heat-tolerance attributes and therefore represent valuable genetic resources for breeding programs targeting terminal heat stress adaptation (Supplementary Table 4).

Although grain yield under terminal heat stress exhibited a significant negative correlation with days to 50% heading (DFH; r = −0.45***), suggesting that earlier flowering may contribute to partial phenological escape, the identification of superior genotypes was not solely driven by earliness. The selected genotypes displayed a wide range of flowering times (56.57–64.47 days), indicating that both early- and late-flowering entries were represented among the superior group. For instance, late-flowering genotypes such as RAJ 4238 (64.19 days), MACS4058 (64.47 days), and 21HTWYT-47 (63.10 days) maintained favorable grain yield and stress tolerance indices under terminal heat stress. In contrast, the check cultivar RG-1 flowered relatively early (58.06 days) but exhibited high susceptibility (SSI = 0.92), the greatest yield reduction (54.03%), and the lowest yield stability index (YSI = 0.46). These findings indicate that genotype classification reflected the ability to sustain productivity under heat stress rather than phenological escape alone. This interpretation is further supported by the strong positive associations of grain yield with spike fertility, thousand-grain weight, and number of grains per spike, highlighting the importance of reproductive resilience and grain filling in conferring terminal heat stress tolerance.

4 Discussion

Terminal heat stress is one of most significant constraints to global heat production, particularly regions where rising temperature coincides with reproductive and grain filling stages. Exposure to high temperatures during these critical developmental phases adversely affects photosynthesis, pollen viability, grain filling, and ultimately grain yield (Farooq et al., 2011; Prasad and Djanaguiraman, 2014). Therefore, identifying novel sources of heat tolerance from genetically diverse germplasm is essential for developing climate-resilient wheat cultivars. The present study evaluated a diverse panel of 500 wheat lines to identify genotypes tolerant to terminal heat stress and to assess the effects of moderate to severe heat stress on key phenological and physiological traits. While our previous study from the same experimental population examined trait relationships using data from a single cropping season (Kumar et al., 2026), the present investigation extends that work by integrating two years of phenotypic data and complementary analytical approaches to identify superior heat-tolerant genotypes. Delayed sowing was used to impose terminal heat stress by exposing plants to elevated temperatures during reproductive and grain-filling stages (Gupta et al., 2015; Nagar et al., 2015; Rehman et al., 2021). Significant genotypic differences were observed for all evaluated traits under heat-stress conditions, indicating substantial genetic variability in stress response. The greater reduction in mean performance and trait range under very late sowing compared with late sowing reflected increased stress severity. Temperature was more than 30 °C during terminal stage in very late sown condition and it called as severe heat stress in wheat (Akter and Rafiqul Islam, 2017). As the impact of high temperature depends on both stress duration and developmental stage, reductions in phenological traits under heat stress serve as useful indicators for identifying heat-tolerant genotypes capable of maintaining developmental stability under elevated temperatures (Rehman et al., 2021; Fu et al., 2023).

The present study clearly demonstrated that the severity of terminal heat stress increased from late sown (LS) to very late sown (VLS) conditions, as reflected by substantially higher percentage reductions across physiological, phenological, and yield-related traits under VLS. Chlorophyll-related traits showed marked sensitivity to heat stress, with SPAD at heading (SPADH) and maturity (SPADM) declining by 13.91% and 22.68% under LS, respectively, but nearly doubling under VLS (26.79% and 47.20%). Similar reductions in SPAD values under terminal heat stress have been widely reported and are attributed to accelerated chlorophyll degradation and impaired photosynthetic efficiency under high-temperature regimes (Rehman et al., 2021; Fu et al., 2023). Canopy temperature (CT) exhibited comparatively lower reductions under both LS (7.47%) and VLS (14.86%), suggesting that genotypic variation in transpirational cooling may partially buffer heat stress effects, as previously reported in wheat under terminal heat conditions (Gautam et al., 2015; Rehman et al., 2021).

Phenological traits were also substantially affected by delayed sowing. Days to 50% heading (DFH), grain filling duration (GFD), and days to physiological maturity (DPM) showed greater reductions under VLS (25.43%, 29.76%, and 27.03%, respectively) compared with LS, reflecting heat-induced acceleration of developmental processes. Shortening of phenological phases under terminal heat stress is a well-documented response in wheat and is primarily driven by temperature-mediated increases in developmental rates, often resulting in reduced assimilate accumulation and lower yield potential (Rehman et al., 2021; Fu et al., 2023). Spike fertility (SF) was particularly sensitive to extreme heat, with reductions increasing from 8.46% under LS to 33.67% under VLS, indicating increased floret sterility and pollen viability loss under high temperatures, as reported earlier (Kumar S et al., 2023; Fu et al., 2023; Gaikwad et al., 2025).

Yield and yield-contributing traits were most severely affected under VLS. Thousand-grain weight (TGW) and number of grains per spike (NGS) declined by 26.25% and 26.45% under LS, respectively, but reductions intensified to 45.30% and 51.49% under VLS. Consequently, grain yield per plant (GYP) exhibited the highest reduction, increasing from 19.83% under LS to 58.81% under VLS. These pronounced yield losses under extended terminal heat stress are consistent with earlier studies demonstrating that high temperatures during grain filling reduce starch synthesis, shorten grain filling duration, and impair source–sink relationships, ultimately leading to sharp declines in grain weight and yield (Kumar H et al., 2023, Asseng et al., 2015);. Overall, the progressive increase in trait reductions from LS to VLS confirms that very late sowing imposes extreme terminal heat stress and provides a stringent environment for identifying wheat genotypes with superior heat tolerance.

Under very late sown conditions, Pearson’s correlation analysis highlighted the key traits contributing to grain yield expression under terminal heat stress. Grain yield per plant (GYP) exhibited strong and positive associations with spike fertility (SF), number of grains per spike (NGS), thousand grain weight (TGW), and SPAD chlorophyll content at heading (SPADH), indicating that both sink strength and photosynthetic capacity play decisive roles in sustaining yield under high-temperature stress. Positive relationships between yield and grain number, grain weight, and spike fertility under terminal heat stress have been consistently reported in wheat and are attributed to improved floret survival, pollen viability, and assimilate partitioning during reproductive and grain-filling stages (Rehman et al., 2021; Kumar S. et al., 2023). Similarly, the positive association between GYP and SPADH suggests that genotypes capable of maintaining higher chlorophyll content under heat stress retain greater photosynthetic efficiency, which supports grain filling under shortened growth duration, as documented in earlier studies (Fu et al., 2023).

In contrast, grain yield showed significant negative correlations with days to flowering (DFH) and days to physiological maturity (DPM), indicating that delayed phenology was disadvantageous under extreme terminal heat stress. This negative association reflects the adaptive advantage of early flowering and maturity under heat-prone environments, enabling partial escape from prolonged exposure to high temperatures during sensitive reproductive stages (Rehman et al., 2021; Kumar H. et al., 2023). The strong positive correlation observed between DFH and DPM further confirms the tight linkage between phenological phases in wheat, where delayed flowering generally leads to extended maturity duration, thereby increasing vulnerability to terminal heat stress (Fu et al., 2023).

Although lower canopy temperature is widely recognized as an indicator of improved heat tolerance through enhanced transpirational cooling and maintenance of plant water status (Gautam et al., 2015; Rehman et al., 2021), the relatively weak correlations observed in the present study suggest that canopy temperature alone was not the primary determinant of yield variation under severe terminal heat stress. Instead, grain yield was more strongly associated with spike fertility, grain number, and grain weight, indicating that maintenance of reproductive function and sink strength played a more decisive role in sustaining productivity under the very late-sown environment. Similar findings were reported by Gautam et al. (2015), who observed that canopy temperature served as a useful physiological indicator of heat tolerance under terminal heat stress, but its association with grain yield was weaker than that of major yield-contributing traits. These findings suggest that canopy temperature should be considered as a complementary selection criterion rather than a standalone indicator of heat tolerance. However, the limited direct association between CT and grain yield under extreme stress conditions indicates that canopy temperature alone may not fully explain yield variation under very late sowing, particularly when stress intensity overwhelms physiological buffering mechanisms.

Overall, these correlation patterns emphasize that under extreme terminal heat stress, yield performance in wheat is primarily governed by traits associated with sink capacity, assimilate retention, and early phenological adjustment, reinforcing their importance as indirect selection criteria for breeding heat-tolerant genotypes.

Principal component analysis effectively captured the multivariate structure of trait variation under terminal heat stress, with the first two principal components explaining 48.69% of the total variability. PC1 was predominantly associated with grain yield per plant, spike fertility, number of grains per spike, thousand-grain weight, and SPAD chlorophyll content, indicating that yield formation under terminal heat stress is primarily governed by sink strength and photosynthetic capacity. Genotypes located in the positive quadrant of PC1 combined superior yield with favorable physiological traits and were therefore identified as heat tolerant. In contrast, negative PC1 scores were associated with longer phenological duration, reflecting heat sensitivity due to prolonged exposure to high temperatures during reproductive stages. PC2 was largely influenced by phenological traits, highlighting the role of phenological adjustment and heat escape mechanisms under severe stress. The clear separation of traits and genotypes across quadrants confirms that terminal heat tolerance is a complex, multi-trait phenomenon, rather than being controlled by a single trait. Similar effectiveness of PCA biplots in identifying key traits and tolerant genotypes under abiotic stress has been reported in wheat and other crops (Aslam et al., 2017; Maqbool et al., 2016; Subramanian et al., 2019). Overall, PCA proved to be a robust tool for genotype differentiation and trait prioritization under terminal heat stress conditions.

Principal component analysis delineated a distinct group of genotypes positioned in the positive quadrant of the biplot, characterized by strong associations with grain yield and yield-contributing physiological traits under terminal heat stress. Similar PCA-based identification of heat-tolerant, high-yielding genotypes has been reported in wheat, where positive PC1 scores were consistently linked with superior performance under stress conditions (Subramanian et al., 2019). To ensure selection robustness, these genotypes were further evaluated for year-to-year reproducibility of grain yield under very late sown (VLS) environments, as yield stability across seasons is considered essential for reliable identification of heat tolerance under field conditions. Based on this criterion, the top 10% of genotypes exhibiting consistent yield performance across years were shortlisted.

From this stable subset, 22 genotypes were selected using multiple stress tolerance indices. Genotypes exhibiting higher mean productivity (MP), stress tolerance index (STI), geometric mean productivity (GMP), yield stability index (YSI), and yield index (YI), along with lower stress susceptibility index (SSI) and tolerance (TOL), were considered superior, consistent with established selection frameworks for stress-prone environments (Gaikwad et al., 2025; Fu et al., 2023; Kumar S. et al., 2023; Puri et al., 2015). The integration of PCA scores with multi-trait performance and stress indices further validated the superiority of these genotypes, indicating their combined ability to maintain high yield potential and favorable physiological performance under terminal heat stress. These elite genotypes provided a robust basis for subsequent clustering and detailed trait profiling.

PCA-based clustering further resolved the 22selected genotypes into two distinct groups, with Cluster II comprising heat-tolerant genotypes and Cluster I representing moderately heat susceptible genotypes. The clear separation of cluster centroids confirmed the robustness of the PCA-based grouping and reinforced the effectiveness of multivariate approaches for classifying genotypes under terminal heat stress. Similar clustering patterns distinguishing tolerant and moderately tolerant groups have been reported in wheat under heat and other abiotic stresses, highlighting the utility of PCA-derived clustering for genotype stratification (Sarkar et al., 2025; Poudel et al., 2025; Baranwal et al., 2024). The top-performing genotypes from the tolerant cluster were further characterized using radar plots, which provided an integrative visualization of their simultaneous performance across multiple yield and physiological traits. The nine tolerant genotypes (SSI ≤ 0.75) exhibited SPAD values at heading and maturity ranging from 33.27–38.41 and 11.65–19.87, respectively, while canopy temperature ranged from 26.22 to 28.11 °C. These genotypes also maintained higher grain yield under stress (12.96–14.53 g plant^-¹), greater yield stability (YSI = 0.56–0.63), and lower yield reductions (36.57–44.28%) compared with the susceptible check RG-1, which recorded only 10.30 g plant^-¹ grain yield under stress, a yield reduction of 54.03%, and a YSI value of 0.46. The relatively high chlorophyll retention observed among the tolerant genotypes suggests improved maintenance of photosynthetic activity during grain filling, while favorable canopy temperature contributed to sustained physiological functioning under heat stress. Together with higher spike fertility and grain number, these traits likely enhanced assimilate production and partitioning, resulting in superior yield stability under severe terminal heat stress. Such multi-trait profiling facilitates the identification of genotypes combining yield stability with favorable physiological responses under stress. Collectively, these genotypes represent strong candidates for deployment in breeding programs targeting improved terminal heat tolerance in wheat.

5 Conclusion

This study demonstrated substantial genetic variation among 500 wheat genotypes evaluated under a very late-sown terminal heat environment under natural field conditions. Very late sowing successfully generated severe terminal heat stress, resulting 59% grain yield reduction and enabling effective discrimination of genotypes with superior performance under these conditions. However, because delayed sowing also modifies phenology, photoperiod, radiation, and crop duration, the observed genotype performance likely reflects the combined effects of stress avoidance and physiological adaptation. Integration of stress tolerance indices with multivariate analyses identified 22 genotypes exhibiting consistently superior performance under the very late-sown terminal heat environment. Spike fertility, grain number, thousand-grain weight, and chlorophyll retention emerged as important traits associated with improved performance under stress, whereas the limited association between canopy temperature and grain yield suggests that canopy temperature alone may not fully explain yield variation under severe terminal heat stress. The moderate reproducibility of grain yield across years further highlights the importance of continued multi-environment evaluation. These findings provide valuable germplasm and trait information for wheat improvement; however, further multi-location and multi-year validation is required before the identified genotypes can be recommended as stable sources of terminal heat stress tolerance.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fagro.2026.1834412/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Jose Ramon Acosta Motos, Catholic University San Antonio of Murcia, Spain

Reviewed by: Rahul Kumar, Agricultural Research Service (USDA), United States

Rehenuma Tabassum, Sylhet Agricultural University, Bangladesh

Anil Patidar, National Bureau of Plant Genetic Resources (ICAR), India

Footnotes

TS, Timely sown; LS, Late sown; VLS, Very late sown; CIMMYT, International Maize and Wheat Improvement Center; SAWSN, Semi-Arid Wheat Screening Nursery; HTWYT, High Temperature Wheat Yield Trial; ESWYT, Elite Spring Wheat Yield Trial; HPYT, HarvestPlus Yield Trial; HLBSN, Helminthosporium Leaf Blight Screening Nursery; PCA, Principal Component Analysis; PC, Principal Component; DFH, Days to 50% Heading; GFD, Grain Filling Duration; DPM, Days to Physiological Maturity; SPADH, SPAD chlorophyll content at Heading; SPADM, SPAD chlorophyll content at Maturity; CT, Canopy Temperature; SF, Spike Fertility; NGS, Number of Grains per Spike; TKW, Thousand-Kernel Weight; GYP, Grain Yield per Plant; MP, Mean Productivity; STI, Stress Tolerance Index; GMP, Geometric Mean Productivity; SSI, Stress Susceptibility Index; TOL, Tolerance Index; YI, Yield Index; YSI, Yield Stability Index; FDR, False Discovery Rate; ^©, Represent check in some graph; °C, Degree centigrade.